HPS: AI buildout moves private credit beyond buyouts

HPS Investment Partners sees the AI infrastructure buildout pulling private credit away from its sponsor-led roots. In a paper titled 'AI and private credit: Assessing risk and finding opportunity,' first reported by Alternative Credit Investor, the BlackRock-owned manager argues that the asset class's historic definition — sponsor-backed leveraged buyouts — is breaking. Large corporate and investment-grade borrowers are arriving with flexible financing needs for data centres, computing capacity, and the technology that supports them. Lenders should judge these loans as they would any corporate borrower, HPS says.

Private credit has matured to where managers work across a broader range of corporate financing needs, HPS writes. The firm calls the AI buildout 'reshaping the private credit opportunity set.' The scale of planned investment invites comparisons with the late-1990s telecom boom. HPS accepts the comparison but not the conclusion that usually follows.

Much of the fiber laid in the late 1990s went dark after the telecom bubble burst. Today's AI infrastructure, HPS notes, often has committed users before construction finishes. That distinction should drive underwriting. Financing these projects, the firm argues, depends less on predicting technology outcomes and more on evaluating counterparty strength. Many transactions are supported by hyperscalers and investment-grade corporate borrowers whose credit profiles add a protective layer.

If that framework holds, AI infrastructure credit is corporate credit with a different label. The question is whether the borrower can pay, not whether the technology works. What HPS leaves unproven is the durability of the committed-user claim. It does not say whether those commitments are priced to cover debt service through a downturn, nor whether the committed users and the borrowers are the same entities.

Direct lenders hoping for a return of buyout flow will not find a substitute here. AI infrastructure loans are a different product — different counterparties, longer construction timelines, and documentation that has to contend with completion risk, cost overruns, and the economics of power supply. HPS's framework gestures at those issues through counterparty strength; the credit agreements will have to be more specific.

The same discipline applies to software, where recent turbulence has bruised private credit portfolios. HPS attributes the volatility to more than AI uncertainty. Businesses financed during a period of elevated valuations, abundant capital, and excessive leverage carried pre-existing weaknesses, it argues. Earlier this year, the apparent AI threat to software prompted a broad selloff and a rise in redemption requests to private credit funds, the article notes.

HPS describes a significant shift toward software fundamentals. In the boom years, investors prioritized growth over profitability, assuming future scale would produce cash flow. Now, HPS argues, lenders focus on whether companies generate enough cash to service debt, and they prefer visible near- and medium-term earnings to aggressive assumptions about the industry's evolution.

This reads as a return to credit basics, not a new framework. The harder test is the loans already sitting on managers' books, written on the old assumptions. The paper does not say how many would survive a fresh look.

The argument lands in a cooling market. Private Credit Daily's latest US direct lending numbers put volume below half the first-quarter pace. BlackRock TCP Capital recently sold nearly half its BDC portfolio into a continuation vehicle. Partners Group closed a $1bn Asian private credit mandate. In this context, an argument that the opportunity set is broadening gives business development teams something to carry into allocator meetings.

The open question is whether the broadening reaches loan documentation. HPS points lenders to near-term earnings visibility, cash flow sufficient to service debt, and counterparties whose balance sheets can absorb bad news. None of those is new. What would be new is applying them consistently to AI infrastructure loans, where the collateral is young and pricing assumptions are untested.

The committed-user argument is central, and the construction cycle will test it before it is finished. If hyperscalers stay on the leases, these loans look like investment-grade credit with a yield premium. If they walk away, the protection HPS describes unwinds. The next few quarters of data-centre leasing will show which case is real.